DOT HS August 2008 Preliminary Evaluation of Advanced Air Bag Field Performance Using Event Data Recorders

Transcription

1 DOT HS August 28 Preliminary Evaluation of Advanced Air Bag Field Performance Using Event Data Recorders This document is available to the public from the National Technical Information Service, Springfield, Virginia 22161

2 This publication is distributed by the U.S. Department of Transportation, National Highway Traffic Safety Administration, in the interest of information exchange. The opinions, findings, and conclusions expressed in this publication are those of the author(s) and not necessarily those of the Department of Transportation or the National Highway Traffic Safety Administration. The United States Government assumes no liability for its content or use thereof. If trade or manufacturers names or products are mentioned, it is because they are considered essential to the object of the publication and should not be construed as an endorsement. The United States Government does not endorse products or manufacturers.

3 REPORT DOCUMENTATION PAGE Form Approved OMB No Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 121 Jefferson Davis Highway, Suite 124, Arlington, VA , and to the Office of Management and Budget, Paperwork Reduction Project (74-188), Washington, DC AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED DOT HS August 28 Final Report April 26 December TITLE AND SUBTITLE. FUNDING NUMBERS Preliminary Evaluation of Advanced Air Bag Field Performance Using Event Data Recorders HS/DG48 6. AUTHOR(S) H. Clay Gabler (Virginia Tech), Craig P. Thor (Virginia Tech), and John Hinch (NHTSA) 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) U.S. Department of Transportation Research and Innovative Technology Administration Volpe Center Cambridge, MA SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) U.S. Department of Transportation National Highway Traffic Safety Administration 12 New Jersey Avenue SE. Washington, DC SUPPLEMENTARY NOTES 8. PERFORMING ORGANIZATION REPORT NUMBER DOT-VNTSC-NHTSA SPONSORING/MONITORING AGENCY REPORT NUMBER DOT HS Organization contracting to RITA/VNTSC: Virginia Tech, Department of Mechanical Engineering, Blacksburg, VA a. DISTRIBUTION/AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE This report is free of charge from the NHTSA Web site at ABSTRACT (Maximum 2 words) This report describes a preliminary evaluation of the field performance of occupant restraint systems designed with advanced air bag features including those specified in the Federal Motor Vehicle Safety Standard No. 28 for advanced air bags, through the use of event data recorders. Although advanced restraint systems have been extensively tested in the laboratory, we are only beginning to understand the performance of these systems in the field. Because EDRs record many of the inputs to the advanced air bag control module, these devices can provide unique insights into the characteristics of field performance of air bags. This research program investigates the feasibility of using EDR data to evaluate advanced air bags. Specifically, this report discusses (1) the development of an expanded EDR dataset based on data retrieved from NASS/CDS 2, SCI, and CIREN in-depth crash investigations, (2) the validation of the accuracy of EDRs in full-scale crash tests, and (3) the feasibility of using EDRs to monitor the performance of advanced air bag restraints in real-world crashes. 14. SUBJECT TERMS 1. NUMBER OF PAGES Event data recorder, traffic crash, air bag PRICE CODE 17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 2. LIMITATION OF ABSTRACT OF REPORT OF THIS PAGE OF ABSTRACT Unclassified Unclassified Unclassified NSN Standard Form 298 (Rev. 2-89) Prescribed by ANSI Std

4 Acknowledgments The authors wish to acknowledge the U.S. Department of Transportation's Volpe National Transportation Systems Center (Volpe Center) and National Highway Traffic Safety Administration for their support of this research effort. Specifically, we would like to thank Marco dasilva of the Volpe Center. Special thanks to the Insurance Institute for Highway Safety and NHTSA for contributing the EDRs data from their crash tests for this study. Thanks to Ford and Toyota for retrieving the data from EDRs that the research team could not download with the Vetronix Crash Data Retrieval System. We also wish to acknowledge the crash test organizations that harvested and collected the NHTSA EDRs for this study: Transportation Research Center (East Liberty, OH), Karco Engineering (Adelanto, CA), MGA Research Corporation (Akron, NY), and CALSPAN (Buffalo, NY). Finally, thanks to Ashley Thompson, our Virginia Tech undergraduate research assistant, for her help in the organization and analysis of the EDR data. iv

5 Table of Contents REPORT DOCUMENTATION PAGE...iii Acknowledgments...iv List of Figures...vi List of Tables...ix 1. Introduction and Background Development of the EDR Dataset EDR Validation in Full-Systems Crash Tests Evaluation of the Field Performance of Advanced Air Bags...2. Conclusions References...44 Appendix A Comparison of Longitudinal Delta V in EDR Data and Crash Test Instrumentation...4 Appendix B Analysis of Frontal NCAP Air Bag Deployment Times...2 Appendix C Computation of Acceleration from EDR Delta V data... v

6 List of Figures Figure 1. Comparison of Vehicle Velocity versus Time Computed from Crash Test Instrumentation with Associated EDR Data (NHTSA Test 62)...9 Figure 2. EDR Recording Duration versus Actual Crash Pulse Duration...12 Figure 3. EDR Longitudinal Delta V versus Crash Test Delta V Over Full Crash Pulse Duration...13 Figure 4. EDR Longitudinal Delta V versus Crash Test Delta V at t=1 ms...14 Figure. Lateral Delta V of 24 Chevrolet Malibu in Frontal Pole Test (IIHS Test CF3)...19 Figure 6. Distribution of Longitudinal Delta V Values in Deployment Events...21 Figure 7. General Area of Damage in Most Harmful Event in Deployment Crashes...22 Figure 8. Number of Impact Events in Each Crash Involving a Frontal Air Bag Deployment as Observed by NASS Investigator...22 Figure 9. Frequency of Deployment Crashes with Multiple Events Involving Longitudinal Delta V Component as Recorded by EDR...23 Figure 1. Distribution of Vehicle Speed Approximately One Second Before Impact in Deployment Events...2 Figure 11. Probability of Deployment of Driver Air Bag by Longitudinal Delta V 26 Figure 12. Distribution of Right-Front Passenger Air Bag Deployment Decisions by Delta V...27 Figure 13. Distribution of Driver Air Bag Dual-Stage, Single-Stage, and Non- Deployments versus Delta V...27 Figure 14. Longitudinal Delta V versus Vehicle Speed Just Before Collision in CAC Deployment Cases...28 Figure 1. First-Stage Deployment Times versus Model Year in Frontal NCAP Tests...29 Figure 16. Cumulative Distribution (%) of Driver First-Stage Air Bag Deployment versus Deployment Time (msec)...3 Figure 17. Case EDR Delta versus Time (msec)...3 Figure 18. Case Differentiated EDR Delta versus Time (msec)...3 Figure 19. Case EDR Delta versus Time (msec)...36 Figure 2. Case Impact With Small Sign and Pole...37 Figure 21. Case Showing Fire Hydrant Damage on Vehicle s Right...37 Figure 22. Frontal Crash Followed by a Rollover in Which Driver Air Bag Deployed, But Passenger Air Bag Did Not Deploy for a Child in the Right- Front Seat (NASS )...4 Figure 23. Frontal Crash in Which Driver Air Bag Deployed, But Passenger Air Bag Did Not in the Presence of an Adult Right-Front Passenger (NASS )...41 Figure 24. NHTSA Test 31-2 Buick Rendezvous (with EDR time shift of -.3s)...4 vi

7 Figure 2. NHTSA Test Chevrolet Colorado (ext.cab) (with EDR time shift of -.7s)...4 Figure 26. NHTSA Test 26-2 Chevrolet Express (with EDR time shift of -.3s)...4 Figure 27. NHTSA Test Pontiac Montana (with no EDR time shift).4 Figure 28. NHTSA Test 26-2 Saturn Ion Side Impact (with no EDR time shift)...4 Figure 29. NHTSA Test Chevrolet Silverado (EDR time shift of -.s)...4 Figure 3. NHTSA Test Chevrolet Uplander (with EDR time shift of -.2s)...46 Figure 31. NHTSA Test Chevrolet Cobalt (with EDR time shift of.s)...46 Figure 32. NHTSA Test Chevrolet Colorado (2-DR) (with no EDR time shift)...46 Figure 33. NHTSA Test Chevrolet Colorado (4-DR) (with EDR time shift of -.s)...46 Figure 34. NHTSA Test Pontiac Grand Prix (4-DR) (with no EDR time shift)...46 Figure 3. NHTSA Test Buick Lucerne CX (with EDR time shift of -.6s)...46 Figure 36. NHTSA Test Chevrolet HHR (with EDR time shift of -.6s)...46 Figure 37. NHTSA Test Chevrolet Impala (with EDR time shift of -.8s)...46 Figure 38. NHTSA Test 2-2 Pontiac G6 (with EDR time shift of -.49s)...47 Figure 39. NHTSA Test Chevrolet Avalanche (with no EDR time shift)...47 Figure 4. NHTSA Test Chevrolet Avalanche (with no EDR time shift)...47 Figure 41. NHTSA Test Buick Lucerne (with EDR time shift of -.8s)...47 Figure 42. NHTSA Test Chevrolet Monte Carlo (with EDR time shift of -.6s)...47 Figure 43. NHTSA Test Cadillac DTS (with EDR time shift of -.7s)...47 Figure 44. NHTSA Test Hummer H3 (with EDR time shift of -.s)...47 Figure 4. NHTSA Test Pontiac G6 (with EDR time shift of -.48s)...47 Figure 46. NHTSA Test Chevrolet Silverado (with EDR time shift of -.6s)...48 Figure 47. NHTSA Test Saturn Aura (with EDR time shift of -.s)...48 Figure 48. NHTSA Test Pontiac Solstice (with EDR time shift of -.s)...48 vii

8 Figure 49. NHTSA Test Chevrolet Silverado (with EDR time shift of -.6s)...48 Figure. IIHS Test CF3-24 Chevrolet Malibu Pole Test (with no EDR time shift)...48 Figure 1. IIHS Test CEF419-2 Saturn Ion (with no EDR time shift)...48 Figure 2. IIHS Test CEF6-2 Chevrolet Colorado (with no EDR time shift)...48 Figure 3. IIHS Test CEF11-2 Buick LaCrosse (with no EDR time shift).48 Figure 4. NHTSA Test Ford (with no EDR time shift)...49 Figure. NHTSA Test Ford Freestyle (with no EDR time shift)...49 Figure 6. NHTSA Test Ford Econoline (with no EDR time shift)...49 Figure 7. NHTSA Test Toyota Camry Side Impact (with no EDR time shift)...49 Figure 8. NHTSA Test Toyota Camry (with no EDR time shift)...49 Figure 9. NHTSA Test 16-2 Toyota Corolla (with no EDR time shift)...49 Figure 6. NHTSA Test 17-2 Toyota Corolla Side Impact (with no EDR time shift)...49 Figure 61. NHTSA Test 29-2 Toyota Matrix (with no EDR time shift)...49 Figure 62. NHTSA Test Toyota Matrix Side Impact (with no EDR time shift)... Figure 63. NHTSA Test Toyota RAV4 (with no EDR time shift)... Figure 64. NHTSA Test Toyota Sienna (with no EDR time shift)... Figure 6. NHTSA Test Toyota Sienna Side Impact (with no EDR time shift)... Figure 66. NHTSA Test Toyota Tacoma (with no EDR time shift)... Figure 67. NHTSA Test Toyota 4Runner (with no EDR time shift).. Figure 68. NHTSA Test Toyota Prius (with no EDR time shift)... Figure 69. NHTSA Test Toyota Tundra (with no EDR time shift)... Figure 7. NHTSA Test Toyota Tundra (with no EDR time shift)...1 Figure 71. NHTSA Test Toyota TC (with no EDR time shift)...1 Figure 72. First Stage Deployment Times versus Model Year In Frontal NCAP Tests...3 Figure 73. Vehicle Longitudinal Deceleration (G s) From an NCAP Test of a 2 Chevrolet Equinox versus Time (msec)... Figure 74. Vehicle EDR Longitudinal Velocity Change (mph) From an NCAP Test of a 2 Chevrolet Equinox versus Time (msec)...6 Figure 7. Differentiated Vehicle EDR Longitudinal Velocity Change (G s) From an NCAP Test of a 2 Chevrolet Equinox versus Time (msec)...6 Figure 76. Differentiated Vehicle EDR Longitudinal Velocity Change Compared With Vehicle Accelerometer (G s) From an NCAP Test of a 2 Chevrolet Equinox versus Time (msec)...7 viii

9 List of Tables Table 1. NHTSA EDR Cases by Database...3 Table 2. Crash Tests Included in the Analysis and the Corresponding EDR... Table 3. Distribution of EDRs in Crash Tests by Test Sponsor and Test Type...7 Table 4. Distribution of EDRs in Crash Tests by Model Year...7 Table. Distribution of EDRs in Crash Tests by Vehicle Make and Air Bag Type...7 Table 6. Summary of EDR Accuracy Based on Comparison With Crash Tests..1 Table 7. Percent Error of the Maximum Longitudinal Delta V for Vehicles That Recorded the Entire Crash Event or Were Missing Less Than 2 percent of the Crash Pulse Data...14 Table 8. Percent Error of the Longitudinal Delta V at 1ms for All Tests...1 Table 9. Accuracy of Pre-Crash Measurements for the EDR and Crash Test...16 Table 1. Accuracy of EDR Driver Belt Buckle Status...18 Table 11. Accuracy of EDR Right Front Passenger Belt Buckle Status...18 Table 12. Distribution of CAC Air Bag Cases by EDR Module Type...21 Table 13. Driver Belt Buckle Status...24 Table 14. Right Front Passenger Belt Buckle Status...24 Table 1. Driver Air Bag Deployment Status...24 Table 16. Vehicle Model Year, Make and Model (* = CAC Vehicle)...3 Table 17. Driver Belt Status, Vehicle Longitudinal Delta V, and Driver Air Bag Deployment Times (* = CAC Vehicle)...31 Table 18. Distribution of Model Years in Delayed Deployment Cases...36 Table 19. Summary of Delayed Deployments (* = CAC Vehicle)...38 Table 2. Frequency of Right-Front Passenger Air Bag Non-Deployments in Crashes in Which the Driver Air Bag Deployed in CAC Vehicles...39 Table 21. Frontal Air Bag Deployment Times by Model Year in NCAP Frontal Crash Tests...2 Table 22. Frontal Air Bag Deployment Times by EDR Module Type in NCAP Frontal Crash Tests...4 ix

10

11 1. Introduction and Background In the United States, automakers have introduced a new generation of advanced occupant restraints in response to the requirements of the Federal Motor Vehicle Safety Standard (FMVSS) No. 28 upgrade (6 FR 368). These advanced systems are characterized by multistage air bag inflators, pretensioners, advanced occupant sensors, and complex air bag deployment algorithms. Air bags in those vehicles certified to the FMVSS No. 28 upgrade are referred to in this report as certified advanced compliant (CAC) air bags. Although these systems have been extensively tested in the laboratory, we are only beginning to understand the performance of these CAC air bags in the field. Because event data recorders (EDRs) record many of the inputs to the advanced air bag control module, these devices can provide unique insights into the performance of air bags in real world crashes. This research program uses EDR data to investigate the feasibility of using EDR data to evaluate advanced air bags. Specifically, this report discusses (1) the development of an expanded EDR dataset, (2) the validation of the accuracy of EDRs in full-scale crash tests, and (3) the feasibility of using EDRs to monitor the performance of advanced air bag restraints in real-world crashes. 1

12 2. Development of the EDR Dataset Objective The objective of this task was to develop a dataset of all available EDR data from NASS/CDS 2-2, Special Crash Investigation (SCI), and CIREN in-depth crash investigation cases. This dataset was the basis for a comprehensive EDR study conducted by researchers at the Volpe Center (dasilva, 28). Approach Our earlier EDR studies have relied on EDR data from NASS/CDS 2-24 (Gabler & Niehoff, 2). The objective of this effort was to expand this EDR dataset to include NASS/CDS 2, Special Crash Investigation (SCI) cases and CIREN cases. NHTSA supplied the research team with all available EDR data for NASS/CDS 2, SCI, and CIREN cases with EDR data. For each case in the dataset, NHTSA crash investigators had downloaded EDR data from the case vehicle using the Vetronix Crash Data Retrieval (CDR) system. At the time of this report, the Vetronix system could only download General Motors (GM) and Ford EDRs. The Vetronix system displays the contents of the EDR, and outputs a CDR file containing the EDR data in binary form. The CDR files are small (typically a few thousand bytes), can be read again by the Vetronix software, are check-summed to prevent tampering, and provide an excellent method for archiving EDR data. CDR files are the most reliable form of EDR data, and were the exclusive basis for our development of the EDR dataset. For each case, the research team read each CDR file using the Vetronix CDR software, converted the EDR data into a usable format, and when possible matched each CDR file with the corresponding crash investigation case. The research team followed a very stringent process in matching the EDR data for each case with the corresponding crash investigation case. For a case to be included in the dataset, we required an exact match of both the case ID (case year, PSU, and case number) and the Vehicle Identification Number (VIN). The name of each CDR file typically contained the case year, PSU, and case number. There was no standard format however for these file names. The CDR file itself contained the full 17-character Vehicle Identification Number (VIN) of the vehicle. The NASS/CDS SAS files only store the first 11 characters of the VIN in order to protect the privacy of the vehicle occupants. Consequently, VIN matching was conducted using the first 11 characters of the VIN. The entire process could only be conducted for NASS/CDS data as only this database was publicly available in a SAS format. This matching method could not be applied to the SCI and CIREN cases as no such SAS files were available. 2

13 Results As shown in Table 1 below, the result was an EDR dataset containing over 2, cases. Table 1. NHTSA EDR Cases by Database Database Years GM Cases Ford Cases Total NASS/CDS 2-2 2, ,32 SCI CIREN Total 2, ,43 Over 9 percent of the cases in the resulting EDR dataset were from NASS/CDS. Almost the entire dataset consists of EDRs from GM vehicles. Only 79 cases in the EDR data were from Ford vehicles. DaSilva (28) presents a comprehensive examination of the contents of the expanded EDR dataset. This description will not be repeated in this report. Rather our report will focus on the validity of EDR data as measured in full systems crash tests, and the use of the expanded EDR dataset to examine the performance of advanced air bags in real world crashes. 3

14 3. EDR Validation in Full-Systems Crash Tests Introduction Before using EDR data to study advanced air bag performance, a key first step was to establish the validity of EDR delta V measurements. Several previous studies have investigated the accuracy of EDRs in earlier model passenger vehicles. The studies can be divided into two groups: low-speed non-deployment evaluations and higher-speed crashes in which the air bag deployed. Accuracy has been found to be very good at the higher speeds (greater than 4 km/hr) typically associated with serious occupant injuries. Chidester et al. (21) investigated the accuracy of GM EDRS in MY 1998-era vehicles. Chidester found that in full systems crash tests the EDRs frequently under reported the delta V by a small amount and that some EDR delta V data was incomplete. The paper did not specify the magnitude of this small error or whether the underestimation of delta V was due to incomplete recording. In a large series of low speed tests, Lawrence et al. (23) also found that GM EDRs understated delta V by a small amount. Comeau et al. (24) examined the EDRs of three different GM vehicle models involved in eight crash tests and reported the tests to have a delta V error of +/- 1 percent. Niehoff et al. (2) examined the accuracy of 37 GM, Toyota, and Ford vehicles of MY 2-2 in a variety of impact scenarios including full frontal, offset frontal, side impact, and vehicle-to-vehicle angled crash tests. The EDRs in this sample were reported to have an average error of 6 percent for the delta V when the entire crash pulse is recorded. Niehoff reported that the EDRs in his sample frequently did not record the entire event. Objective The objective of this section is to examine the accuracy of EDR data in a range of crash test scenarios for model year cars and light trucks. Approach This study examines the accuracy of EDR data downloaded from 48 crash-tested vehicles. Our approach was to evaluate the accuracy of the EDR data by comparison with the corresponding lab-grade instrumentation data from the crash test. Crash tests from both the National Highway Traffic Safety Administration (NHTSA) and the Insurance Institute for Highway Safety (IIHS) were included. Approximately two-thirds of the EDRs in the sample (31 of 48) were from CAC-equipped restraint systems. Table 2 details the crash tests used in this analysis including make, model, crash configuration, impact speed, and testing organization. 4

15 Table 2. Crash Tests Included in the Analysis and the Corresponding EDR Test Number Vehicle Description Test Performer Test Type Closing Speed (mph) Impact Angle (deg) Test Offset (%) Barrier CAC EDR Model 31 2 Buick Rendezvous NHTSA Full Frontal 34.9 Rigid Y SDMDW Chevrolet Colorado NHTSA Full Frontal 3.2 Rigid N SDMGF Chevrolet Express NHTSA Full Frontal 34.9 Rigid N SDMGF Pontiac Montana NHTSA Full Frontal 34.8 Rigid Y SDMDW Saturn Ion NHTSA Side Impact MDB N SDMDW Chevrolet Silverado NHTSA Full Frontal 34.9 Rigid Y SDMGF Chevrolet Uplander NHTSA Full Frontal 34.9 Rigid Y SDMDW Chevrolet Cobalt NHTSA Full Frontal 34.9 Rigid N Epsilon Chevrolet Colorado NHTSA Full Frontal 34.9 Rigid Y SDMGF Chevrolet Colorado NHTSA Full Frontal 34.8 Rigid Y SDMGF Pontiac Grand Prix NHTSA Full Frontal 3.1 Rigid N SDMDW Buick Lucerne CX NHTSA Full Frontal 3.1 Rigid Y SDMC Chevrolet HHR NHTSA Full Frontal 34.9 Rigid Y Epsilon Chevrolet Impala NHTSA Full Frontal 3.2 Rigid Y SDMC Pontiac G6 NHTSA Full Frontal 3.3 Rigid N Epsilon Chevrolet Avalanche NHTSA Full Frontal 3. Rigid Y SDMGF Chevrolet Avalanche NHTSA Full Frontal 3.1 Rigid Y SDMGF Buick Lucerne NHTSA Full Frontal 24.7 Rigid Y SDMC Chevrolet Monte Carlo NHTSA Full Frontal 3. Rigid N SDMC Cadillac DTS NHTSA Full Frontal 3.2 Rigid Y SDMC Hummer H3 NHTSA Full Frontal 3. Rigid Y SDMDS Pontiac G6 NHTSA Full Frontal 24.7 Rigid N Epsilon Chevrolet Silverado NHTSA Full Frontal 3.1 Rigid Y SDMC Saturn Aura NHTSA Full Frontal 3.1 Rigid Y Epsilon Pontiac Solstice NHTSA Full Frontal 3. Rigid Y Epsilon Chevrolet Silverado NHTSA Full Frontal 34.8 Rigid Y SDMC26 CF3 24 Chevrolet Malibu IIHS Pole Pole N Epsilon

16 Test Number Vehicle Description Test Performer Test Type Closing Speed (mph) Impact Angle (deg) Test Offset (%) Barrier CAC EDR Model CEF419 2 Saturn Ion IIHS Frontal Offset 4. 4 Deformable N N/A CEF6 2 Chevrolet Colorado IIHS Frontal Offset Deformable N N/A CEF11 2 Buick LaCrosse IIHS Frontal Offset 4. 4 Deformable Y N/A Ford NHTSA Full Frontal 3.3 Rigid Y N/A Ford Freestyle NHTSA Full Frontal 3.1 Rigid Y N/A Ford Econoline NHTSA Full Frontal 3. Rigid N N/A Toyota Camry NHTSA Side Impact MDB Y Toyota Camry NHTSA Full Frontal 3.1 Rigid Y *4-( ) 16 2 Toyota Corolla NHTSA Full Frontal 3.1 Rigid Y Toyota Corolla NHTSA Side Impact MDB Y Toyota Matrix NHTSA Full Frontal 3.1 Rigid Y Toyota Matrix NHTSA Side Impact MDB Y Toyota RAV4 NHTSA Full Frontal 3.3 Rigid N Toyota Sienna NHTSA Full Frontal 3. Rigid Y Toyota Sienna NHTSA Side Impact MDB N Toyota Tacoma NHTSA Full Frontal 34.9 Rigid Y Toyota 4Runner NHTSA Full Frontal 34.9 Rigid N Toyota Prius NHTSA Full Frontal 3.4 Rigid N Toyota Tundra NHTSA Full Frontal 3. Rigid Y 8917-C Toyota Tundra NHTSA Full Frontal 3. Rigid Y 8917-C Toyota Scion TC NHTSA Full Frontal 34.9 Rigid N Note: MDB = Movable Deformable Barrier 6

17 Results Table 3 provides a summary of the types of tests from which EDR data was downloaded. The tests were conducted at test speeds ranging from 2 mph to 4 mph. The majority of tests (39 of 48) were full-frontal, rigid-barrier tests performed primarily at 3 mph. The sample also included three 4 percent offsetfrontal deformable barrier tests conducted at 4 mph and one 1 percent offset pole test performed at 4 mph. In the five side impact tests, the subject vehicle was struck by a moving deformable impactor at a speed of 38 mph. EDRs for this study were downloaded from General Motors, Toyota, and Ford vehicles as shown in Table 3. Model years ranged from 24 to 27 as shown in Table 4. Most of the EDRs in the sample were from GM vehicles (3 of 48). Over half of the vehicles were from model year 2 (2 of 48). As shown in Table, approximately two-thirds of the EDRs in the sample (31 of 48) were from CAC-equipped restraint systems. Table 3. Distribution of EDRs in Crash Tests by Test Sponsor and Test Type Agency Test Type Impact Speed Number of Cases (mph) Full Frontal Rigid Barrier 3 37 Full Frontal Rigid Barrier 3 1 NHTSA Full Frontal Rigid Barrier 2 1 Side Impact 38 4% Frontal Offset 4 3 IIHS 1% Offset Pole 4 1 Total 48 Table 4. Distribution of EDRs in Crash Tests by Model Year Model Year Number of Cases Total 48 Table. Distribution of EDRs in Crash Tests by Vehicle Make and Air Bag Type Vehicle Make Non-CAC CAC Total General Motors Toyota 1 1 Ford Total

18 GM EDRs from NHTSA tests were downloaded by the research team using the Vetronix Crash Data Retrieval System. IIHS downloaded and provided the EDR data from four of their tests for this study. The EDRs from the Ford and Toyota tests included in this analysis were harvested from the vehicle and sent to the respective companies to be downloaded. The GM EDRs recorded delta V every 1 ms for recording durations ranging from 1-24 ms. The Ford EDRs recorded both acceleration and delta V every 1 ms for durations up to 29 ms. One model of the Toyota EDR was observed to record every 1 ms for up to 1 ms, while a second Toyota EDR model was observed to record at 1.2 ms intervals for a 14 ms duration. The crash test data for comparison with the EDR data was obtained from the NHTSA and IIHS crash test databases. The NHTSA crash test data was analyzed using the NHTSA Signal Browser software. Accelerometers mounted within the occupant compartment of each vehicle were selected for comparison with the EDR. The EDR is located within the passenger compartment, often under the front seat or in the center console. All comparisons used accelerometers aligned with the longitudinal axis of the vehicle. An assessment of EDR lateral delta V was not possible as none of the EDRs in side-impact tests recorded lateral delta V. Vehicle acceleration data for the IIHS tests was accessed through the IIHS Tech Data site. The IIHS database did not identify the location of the sensor in the crash. Time-Shifting of EDR Delta V Data Air bag deployment is controlled using a microprocessor. Typically vehicle acceleration, often measured at a central vehicle location and near the front of the vehicle, is processed to determine when the vehicle s frontal air bags should be deployed as well as which air bag stage should be used. The air bag processor wakes up after it senses a predetermined acceleration threshold has been exceeded. This wake-up is defined as algorithm enable (AE) (Chidester et al, 1999). After AE occurs, the processor continues to monitor and analyze the vehicle s deceleration profile and determines if and when the air bags should be deployed. The time the processor deploys the air bags is often referred to as air bag deployment time and is referenced to AE as a time zero. For instance, if the air bags deployed 2 milliseconds (msec) after AE, common notation would consider this an air bag deployment time of 2 msec. For an EDR, time zero is the time of algorithm enable or algorithm wakeup. Algorithm enable typically occurs only after the EDR has measured 1-2 G s deceleration typically a few milliseconds after impact in a frontal barrier crash. One consequence of this recording delay is that because algorithm enable does not happen immediately, a small change in velocity typically 1 to 2mph is not recorded. Finally, GM EDRs record for up to ms prior to the air bag triggering. 8

19 The crash test data and the EDR data were overlaid on a plot for qualitative comparison. As shown in Figure 1, time zero for the EDR records frequently did not coincide with time zero for the crash test instrumentation. A time shift of the EDR data was required to allow comparison with the crash test instrumentation. The time shifting was conducting manually for each test by visually aligning the EDR data with the crash test data. Improved time shifting may be possible using a numerical technique such as that developed by Niehoff (2). A vehicle velocity versus time plot containing both the EDR and crash test curves were created for each of the crash tests and are presented in the appendices Longitudinal - Right Rear Sill (ch. 9) Time Shift Figure 1. Comparison of Vehicle Velocity versus Time Computed from Crash Test Instrumentation with Associated EDR Data (NHTSA Test 62) Crash Pulse Duration Each of the crash tests was analyzed to determine whether the entire crash pulse was recorded. Most crash pulses in our sample had duration of approximately 1-1 ms. Crash pulse duration was defined to be the time interval between the time of initial impact and the time of maximum delta V. The length of the crash pulse is a strong function of the crash test type. An IIHS 4 percent offset crash test, for example, can last over 2 ms while an NCAP full frontal rigid barrier crash may only last 1-12 ms. Table 6 reveals that the longest crash pulse for a longitudinal impact was indeed an IIHS offset test (24 ms). Hence, an EDR that only has a recording time of 1-1 ms may be missing a large portion of the crash information. 9

20 Table 6. Summary of EDR Accuracy Based on Comparison with Crash Tests Test Number EDR Max Delta V (mph) Crash Test Max Delta V (mph) Max Delta V Error (%) EDR Delta ms (mph) Crash Test Delta ms (mph) Delta V 1ms (%) Lateral Delta V Recorded (Y/N) EDR Time Shift (ms) EDR Recording Time (ms) Crash Pulse Duration Estimated (ms) Crash Pulse Duration Error (%) % % % % % None % % % % % % % % None % % % % % % % % % % % % % % % % % Y None % % Y None % % Y None % % Y None % % % % % % % % Y % % % Y None % % Y None % % Y None % % Y None % % Y None % % Y None 1

21 Test Number EDR Max Delta V (mph) Crash Test Max Delta V (mph) Max Delta V Error (%) EDR Delta ms (mph) Crash Test Delta ms (mph) Delta V 1ms (%) Lateral Delta V Recorded (Y/N) EDR Time Shift (ms) EDR Recording Time (ms) Crash Pulse Duration Estimated (ms) Crash Pulse Duration Error (%) % % Y None % % Y None % % Y 1 14 None CEF % % Y None CEF % % % CEF % % % CEF % % % % % Y 29 9 None % % Y None % % Y None % % None % % None % % None % % None % % None % % None % % None % % None % % None % % None % % None % % None % % None % % None % % None 11

22 As shown in Figure 2, 14 of 48 EDRs (29.2%) did not record the entire event. This is an improvement however over the findings of the Niehoff study that reported that the majority of the EDRs in its samples did not successfully record the entire event. The worst case was IIHS frontal offset test CEF6 of a 2 Chevrolet Colorado. In this test, only the first 128 ms of the 24-ms-long crash pulse was recorded by the EDR. 2 EDR Recording Time (ms) All data recorded Some data lost Crash Pulse Duration (ms) Figure 2. EDR Recording Duration versus Actual Crash Pulse Duration It should be noted that maximum crash pulse recording duration is simply a function of the amount off computer memory onboard each EDR. The older GM EDRs in our dataset had sufficient memory to store only up to 1 ms of the crash pulse. The GM algorithm in these modules however called for only storing up to 1 ms after air bag triggering. This constraint led to many of the pulse durations below 1 ms. We note that the latest generation GM EDRs can now store up to 3 ms of the crash pulse. We anticipate that this ability to store longer crash pulses will be extended to other GM models as the price of computer memory continues to drop. Accuracy of EDR Delta V measurements EDRs that do not record the entire event will underestimate the delta V not because of sensor inaccuracy, but because of recording capacity. To get a better measure of measurement accuracy, we first restricted our analysis to those tests for which the EDR recorded the entire crash pulse or were missing no more than 2 percent of the crash data. In these cases, the EDRs were 12

23 successful at recording with a significant amount of accuracy as compared to the test instrumentation. As shown in Figure 3, EDR delta V underestimates true delta V by under. percent on average for crash pulses that were completely recorded by the EDR. The correlation between EDR delta V and true delta V for this dataset is very high with R 2 =.972. It should be noted that this dataset of complete EDR recordings contains full frontal barrier crash tests, the longitudinal delta V component of side-impact crash tests and a single frontal pole tests. The dataset does not include any frontal-offset crash tests. EDR Longitudinal Delta-V at 1 ms (mph) y =.992x R 2 = Crash Test Longitudinal Delta-V at 1 ms (mph) Figure 3. EDR Longitudinal Delta V versus Crash Test Delta V Over Full Crash Pulse Duration To compare the accuracy of the EDR for all tests, rather than just those tests in which the entire event was recorded, the EDR delta V and crash test delta V were next compared t=1 ms. All EDRs in our dataset recorded at least 1 ms. The analysis at t=1 ms from the EDR included the time shift for each respective test to ensure the point of impact matches for both the EDR and the crash test data. As shown in Figure 4, EDR delta V underestimates true delta V by under. percent on average for crash pulses at t=1 ms. The correlation between EDR delta V and true delta V for this dataset is very high with R 2 =.964. This analysis included all crash tests in our dataset including the crash pulses from full frontal, frontal-offset and frontal-pole crash tests as well as the longitudinal component of crash pulses from side impact tests. 13

24 EDR Longitudinal Delta-V at 1 ms (mph) y =.994x R 2 = Crash Test Longitudinal Delta-V at 1 ms (mph) Figure 4. EDR Longitudinal Delta V versus Crash Test Delta V at t=1 ms Accuracy based on Average Absolute Percent Error Niehoff (2) used a different technique to compute EDR accuracy. His approach was to compute the percent difference between the EDR delta V and true delta V for each test, and then average the absolute values of each percent difference. This approach gives a very conservative estimate of EDR error. We repeat the approach on this new dataset for comparison with the Niehoff (2) results. Table 7 shows that the average absolute percent difference between EDR and crash test delta V was 4.2 percent. This error is a slight improvement over the 6 percent error reported by Niehoff et al. (2). Using the Niehoff approach, all averages presented were based on the absolute value of the percent error. Table 7. Percent Error of the Maximum Longitudinal Delta V for Vehicles That Recorded the Entire Crash Event or Were Missing Less Than 2 percent of the Crash Pulse Data Frontal Side Impact Impact Fraction of Crash Pulse Longitudinal Longitudinal Duration Unrecorded All Delta V Delta V N Average loss 4.2% 3.% 1.8% Standard Deviation.8% 4.1% 12.7% Minimum loss.%.%.% Maximum loss 28.6% 21.8% 28.6% 14

25 The maximum longitudinal delta V was predicted more accurately in the front tests than in the side impact tests. The side impact tests in our sample had a fairly low longitudinal delta V. When compared in absolute terms instead of percent error, the lateral tests had an average error of only. mph for the longitudinal delta V. Unfortunately, none of the EDRs with lateral delta V data were subjected to lateral impact crash tests; therefore, that analysis could not be included in this study. The average absolute value percent error for longitudinal delta V at 1-ms for the crash tests in our dataset are presented in Table 8. Table 8. Percent Error of the Longitudinal Delta V at 1ms for All Tests All Tests Front Impact Tests Side Impact Tests - Full Barrier 4% Offset Pole All Front All N Average loss 7.% 4.3% 3.6% 7.2% 4.3% 3.3% Standard Deviation 12.6% 4.2%.3% - 4.1% 27.7% Minimum loss.%.% 3.2% 7.2%.%.% Maximum loss 82.6% 22.6% 4.% 7.2% 22.6% 82.6% The EDR was able to predict the longitudinal delta V in full frontal, frontal-offset, and pole tests with reasonable accuracy. The side impacts did not show the same level of accuracy as the longitudinal tests on a percentage basis, but the average in absolute terms had an error of only.7 mph for the lateral impact tests at 1ms. Accuracy of EDR Pre-Crash Speed The pre-crash vehicle speed in our sample was evaluated for accuracy by comparison with the known crash test impact speeds. Table 9 shows that the pre-crash speed of the vehicle as recorded by the EDR was always within 3 percent with the exception of test 31. In test 31, the EDR underreported the pre-crash speed by 22 percent. The EDR download information provided by Toyota did not provide non-zero, pre-crash vehicle speed for any case except test 269. It is not known if this is a result of the EDR recording capabilities or simply an artifact of the downloading method. Both tests 269 and 31 were New Car Assessment Program (NCAP) full frontal rigid barrier crash tests. 1

26 Table 9. Accuracy of Pre-Crash Measurements for the EDR and Crash Test EDR Pre- Actual Crash Pre-Crash Driver Belt Passenger Vehicle Vehicle Status Belt Status Test Speed Speed Number Vehicle Description EDR Test EDR Test (mph) (mph) % Error 31 2 Buick Rendezvous Y Y - Y % Chevrolet Colorado (ext.cab) Y Y - Y % 26 2 Chevrolet Express Y Y - Y % Pontiac Montana Y Y - Y % 26 2 Saturn Ion Y Y - N. % Chevrolet Silverado (crew cab) Y Y Y Y % Chevrolet Uplander Y Y - Y % Chevrolet Colorado (2-DR) Y Y Y Y % Chevrolet Colorado (4-DR) Y Y Y Y % Pontiac Grand Prix (4-DR) Y Y - Y % Buick Lucerne CX Y Y Y Y % Chevrolet HHR Y Y Y Y % Chevrolet Impala Y Y Y Y % 2 2 Pontiac G6 Y Y - Y % Chevrolet Avalanche Y Y Y Y 3 3. % Chevrolet Avalanche N N N N % Buick Lucerne N N N N % Chevrolet Monte Carlo Y Y Y Y % Cadillac DTS Y Y Y Y % Hummer H3 Y Y Y Y % Pontiac G6 N N N N % Chevrolet Silverado Y Y Y Y % Saturn Aura Y Y Y Y % Pontiac Solstice Y Y Y Y % 16

27 Driver Belt Status Passenger Belt Status Test Number Vehicle Description EDR Test EDR Test EDR Pre- Crash Vehicle Speed (mph) Actual Pre-Crash Vehicle Speed (mph) % Error Chevrolet Silverado Y Y Y Y % Chevrolet Cobalt Y Y - Y % CF3 24 Chevrolet Malibu - Y - Y % CEF419 2 Saturn Ion - Y - N 4 4. % CEF6 2 Chevrolet Colorado Y Y - N % CEF11 2 Buick LaCrosse Y Y - N % Ford Y Y Y Y % Ford Freestyle Y Y Y Y % Ford Econoline - Y - Y % Toyota Camry Y Y N N Toyota Camry Y Y Y Y Toyota Corolla Y Y Y Y Toyota Corolla Y Y N N Toyota Matrix Y Y Y Y Toyota Matrix Y Y N N Toyota RAV4 Y Y Y Y Toyota Sienna Y Y Y Y % Toyota Sienna Y Y N N Toyota Tacoma Y Y Y Y Toyota 4Runner Y Y Y Y Toyota Prius Y Y Y Y Toyota Tundra Y Y Y Y Toyota Tundra Y Y Y Y Toyota Scion TC Y Y Y Y

28 Accuracy of EDR Belt Buckle Status Table 9 compared the seat belt buckle status used in each test with the belt buckle status recorded by the EDR. As shown in Table 1, the EDR record of driver belt status was available for 4 of 48 tests. Forty-two drivers were belted and 3 were unbelted. In all cases, the EDR correctly recorded the driver buckle status.. Table 1. Accuracy of EDR Driver Belt Buckle Status Actual Belt EDR Belt Status Total Buckle Status Buckled Unbuckled NA Buckled Unbuckled 3 3 Total Right-front passenger belt buckle status is a relatively new feature of EDRs. As shown in Table 11, belt buckle status was recorded in 36 of the 48 tests in our sample. In all cases, the EDR correctly recorded the RF passenger buckle status. Table 11. Accuracy of EDR Right Front Passenger Belt Buckle Status Actual Belt EDR Belt Status Total Buckle Status Buckled Unbuckled NA Buckled Unbuckled No Passenger 3 Total

29 Comparison of Lateral Delta V Accuracy in EDRs As shown in Table 6, 18 of the 48 vehicles in our EDR sample had the capacity to record lateral delta V in addition to longitudinal delta V. Only 1 vehicle, a 24 Chevrolet Malibu subjected to a frontal pole test (IIHS Test CF3) had a nonzero record of lateral delta V. Figure compares the lateral delta V as recorded by the EDR with lateral delta V as measured by crash test instrumentation v (mph)_iihs Figure. Lateral Delta V of 24 Chevrolet Malibu in Frontal Pole Test (IIHS Test CF3) Agreement between the EDR and crash test instrumentation is reasonably good for the first ms of the event. However agreement is poor after ms. Niehoff (2) made a similar observation about the accuracy of the lateral delta V recorded by a 24 Chevrolet Malibu subjected to a side impact. Because our sample contained only one EDR with a non-zero lateral delta V, it is unknown whether this finding generalizes to later model vehicles with newer generations of EDRs. 19

30 4. Evaluation of the Field Performance of Advanced Air Bags Introduction In the United States, automakers have introduced a new generation of advanced occupant restraints, including those specifically introduced in response to the requirements for advanced air bags, as specified in the FMVSS No. 28 upgrade (49 CFR [6FR368]). These advanced systems are characterized by multi-stage air bag inflators, pretensioners, advanced occupant sensors, and complex air bag deployment algorithms. Although these systems have been extensively tested in the laboratory, we are only beginning to understand the performance of these systems in the field. Because EDRs record many of the inputs to the advanced air bag control module, these devices can provide unique insights into the performance of air bags in the field. Objective The objective of this study was to characterize the performance of advanced frontal air bags in real-world crashes. The study included both vehicles certified to the FMVSS No. 28 advanced air bag regulation, and vehicles having dualstage frontal air bags. Approach The analysis was based upon EDR records extracted from the NHTSA EDR dataset. NHTSA now has the records from over 2,2 EDRs downloaded as part of National Automotive Sampling System/Crashworthiness Data System (NASS/CDS) 2-2 crash investigations. All cases were downloaded by NASS investigators in the field using the Vetronix crash data recorder retrieval system. Characterization of Dataset This study included only EDR cases from vehicles having a dual-stage frontal air bag. The resulting sample contained the EDR records from 16 vehicles having air bags of the advanced type, also referred to as certified advanced compliant (CAC) air bags. CAC air bags are defined as air bags in those vehicles certified to the FMVSS No. 28 upgrade. The sample was composed entirely of GM passenger cars, light trucks, and vans. Table 12 shows the distribution of cases by EDR module type. 2

31 Table 12. Distribution of CAC Air Bag Cases by EDR Module Type EDR Module Type Deployment Non-Deployment Total SDMDW SDMGF Total GM EDRs record longitudinal delta V versus time for up to two events. Figure 6 presents the distribution of maximum longitudinal delta V recorded by each of 47 the CAC EDRs in which the frontal air bag deployed. The median longitudinal delta V in our sample was approximately 1 mph. 1% 9% 8% Cumulative Frequency (%) 7% 6% % 4% 3% 2% 1% Deployments % EDR Longitudinal Delta- Figure 6. Distribution of Longitudinal Delta V Values in Deployment Events As shown in Figure 7, a frontal impact was the most harmful event in over 9 percent of the CAC air bag deployment cases. 21

32 1% 9% 91% 8% 7% Frequency (%) 6% % 4% 3% 2% 1% % 6% 2% Front Right Side Top GAD in Most Harmful Event Figure 7. General Area of Damage in Most Harmful Event in Deployment Crashes More useful than knowing the general area of damage (GAD) of the most harmful event however would be to know the GAD of the event that triggered the air bag. The most harmful event may not be the event that triggers the air bag. Unfortunately, in a multiple-event crash, the event that triggered the air bag cannot always be determined. As shown in Figure 8, NASS investigators recorded that approximately half of the CAC air bag deployment cases involved multiple events. Not all these events necessarily have a longitudinal component of sufficient magnitude to deploy the air bag. 6% % 4% Frequency (%) 3% 2% 1% % 1 2 Number of Events 3 Figure 8. Number of Impact Events in Each Crash Involving a Frontal Air Bag Deployment As Observed by NASS Investigator 22

33 The EDR data indicated that the majority of the deployment cases in our sample involved only a single event having a longitudinal component of delta V. The SDMGF22 module records a count of the number of events in each crash that involved a longitudinal component of delta V. In our sample of 47 deployments, 44 were SDMGF22 modules. Figure 9 below shows that in over 8 percent of the cases, the EDR detected only a single impact with any longitudinal component. This observation does not however mean these events were frontal impacts. Although events with strong longitudinal components are typically frontal impacts, it is possible for other crash modes including side impacts to have a significantly severe longitudinal component to deploy the air bag. 9% 8% 7% 6% Frequency (%) % 4% 3% 2% 1% % No Yes Figure 9. Frequency of Deployment Crashes with Multiple Events Involving Longitudinal Delta V Component as Recorded by EDR Belt Use and Air Bag Deployment Table 13 presents the distribution of driver belt buckle status in deployment cases. In approximately half of these real-world crashes, the EDR recorded that the driver s seat belt was buckled. In our sample, the EDR driver seat belt buckle status frequently did not agree with the belt use status determined by the NASS investigator. In 9 of the 31 cases in which NASS investigators believed that the driver was belted, the EDR recorded that the driver belt was unbuckled. Note that this finding is in sharp contrast to our observation of EDRs downloaded from late-model crash tests. In crash tests, driver and passenger belt buckle status either buckled or unbuckled was correctly recorded by the EDRs in all cases for which seat belt buckle status was available. 23

34 Table 13. Driver Belt Buckle Status EDR Buckle NASS NASS - Not Total Status Belted Unbelted Inspected by NASS Buckled Not Buckled Total Table 14 shows that in half of the cases in which a right-front passenger was present, the EDR recorded that the passenger was buckled. Because the EDR passenger buckle status is not a data element recorded by the SDMDW23 module, the three SDMDW23 cases are not tabulated in Table 14. Table 14. Right Front Passenger Belt Buckle Status EDR Buckle Status NASS Belted NASS - Unbelted Total Buckled Not Buckled Total Table 1 compares the records of driver air bag deployment as indicated by the NASS investigator and recorded by the EDR. In all but one of the deployments, the EDR and NASS investigators agreed the air bag deployed. In all nondeployment cases, EDR and NASS investigators agreed that the bag had not deployed. EDR Deployment Status Table 1. Driver Air Bag Deployment Status NASS- Not Inspected NASS- Bag Deployed NASS- No Deploy Total Deployed Non-deploy Total In case , a 23 Chevrolet Suburban was involved in a crash in which the EDR recorded that the air bag controller commanded the driver air bag to deploy. However, NASS investigators observed that the driver air bag did not deploy. Inspection of the photos from the investigation confirms the NASS observation that the bag did not deploy. The EDR recorded that the Chevrolet Suburban experienced a longitudinal delta V of 12 mph in this crash. 24

35 Vehicle Speed Just Prior to Impact The GM EDRs in our dataset recorded seconds of pre-crash data in one second intervals for vehicle speed, engine speed, engine throttle setting, and brake status. The vehicle speed data at one second before algorithm enable provides an estimate of vehicle speed approximately one second before impact. Figure 1 provides a distribution of vehicle speed at t = - 1 second for the CAC deployment cases in our sample. Although the EDRs in our dataset did not record impact speed, this measure provides an estimate of vehicle speed just before impact. The median vehicle speed approximately 1 second before impact was 38 mph. 1% 9% 8% Cumulative Frequency (%) 7% 6% % 4% 3% 2% 1% % Vehicle Speed (t = -1 sec) Figure 1. Distribution of Vehicle Speed Approximately One Second Before Impact in Deployment Events RESULTS Figure 11 compares the distribution of the driver air bag deployments and nondeployments by peak longitudinal delta V. All cases in this analysis had incurred a frontal impact in the most harmful event. The cases were aggregated into three groups: (1) those crashes that resulted in a deployment, (2) those crashes not sufficiently severe to deploy the air bag, and (3) split deployments. Split deployments are those cases in which the driver air bag deployed, but the right front passenger air bag did not deploy despite the presence of a passenger. There were no cases in which the passenger air bag deployed, but the driver air bag did not deploy. Of the 16 CAC cases, there were 41 deployments, 2 split deployments, and 19 non-deployments in which the general area of damage was frontal. The driver frontal air bag was observed to deploy in crashes having a longitudinal delta V as low as 3-4 mph. The driver bag was observed to not deploy in a crash 2

36 having a longitudinal delta V of 26 mph. This crash was a long duration crash of approximately 27 milliseconds into an earth and rock embankment. Logistic regression was performed to determine the probability of driver air bag deployment as a function of longitudinal delta V. For this sample, the probability of driver air bag deployment was percent for a longitudinal delta V of 8 mph. 1 Probability of Deployment.7..2 Non-deployment Deployment Split Deployment Prob. of Deployment RF Passenger Present, but RF Passenger Air Bag did not deploy Longitudinal EDR Delta- Figure 11. Probability of Deployment of Driver Air Bag by Longitudinal Delta V In our dataset of 16 CAC cases, there were 2 right front passengers involved in a crash in which a frontal impact was the most harmful event. This 2-case set consisted of 11 deployments, 2 split deployments, and 7 non-deployments. Figure 12 presents the distribution of the right-front air bag deployment decision by longitudinal delta V for these cases. The right-front passenger air bag was observed to deploy in collisions having a longitudinal delta V as low as 6 mph. In general, the passenger air bag did not deploy in low-delta-v crashes. In one crash however, the right-front passenger air bag did not deploy in a crash having a longitudinal delta V of 26 mph. Because our dataset contained only a limited number of right-front passenger cases, a logistic regression computation was not possible for this data subset. 26

37 Non-deployment Deployment Split Deployment Driver Bag Deployed, but not RF Passenger Air Bag Longitudinal EDR Delta- Figure 12. Distribution of Right-Front Passenger Air Bag Deployment Decisions by Delta V All CAC air bag systems in our dataset contained dual-stage inflators. Dualstage inflators allow the air bag deployment characteristics to be tailored to the particular crash severity and / or occupant configuration of a collision (including belt usage). Of the16 CAC cases, there were 43 driver air bag deployments and 19 non-deployments in which the most harmful event was a frontal impact. In the 43 deployments, both the first and second stage fired in 9 of the crashes. Only the first stage fired in the remaining 34 cases. In general as shown in Figure 13, both inflator stages were triggered only in higher delta V crashes. Non-deployment Single Stage Deployment Dual Stage Deployment Longitudinal EDR Delta- Figure 13. Distribution of Driver Air bag Dual-Stage, Single-Stage, and Non-deployments versus Delta V 27

38 Figure 14 presents the relationship between longitudinal delta V and the vehicle speed just prior to impact. In the majority of cases, vehicle speed greatly exceeds longitudinal delta V. 1 9 EDR Longitudinal Delta Vehicle Speed at t=-1 second (mph) 1 Figure 14. Longitudinal Delta V versus Vehicle Speed Just Before Collision in CAC Deployment Cases Time Interval from Algorithm Enable to Deployment To provide context for real world air bag deployment times, EDRs have been used to assess that air bag deployment times during NHTSA s frontal barrier tests, conducted for FMVSS No. 28 and New Car Assessment Program (NCAP). Data from 29 GM vehicles with dual stage inflators, model year 22 through 26, were examined. Details on these crash tests and their associated air bag deployments times are provided in appendix B. Note that this dataset is restricted to frontal NCAP tests of GM vehicles. Many of the crash tests in this dataset were also examined in our analysis of EDR data validity in crash tests presented in an earlier chapter. First-stage deployment times are shown in Figure 1. For the crash tests in this sample, the average deployment time for the first-stage driver air bag was 7 msec, with a range of 2. to 17. msec. Generally, the driver and right-front passenger air bags (both first and second stages) were triggered at exactly the same time. 28

39 1st Stage Deployment Time (ms) Model Year Figure 1. First-Stage Deployment Times versus Model Year in Frontal NCAP Tests Analyses of air bag deployments from real-world crashes would allow full range analysis of deployment times under many circumstances. Since there were only 47 CAC deployment cases, we extended the analysis of deployment times to include pre-cac vehicles with dual stage air bags. NASS cases from years 2 to 2, which included a complete EDR record, and a GM vehicle with a dualstage air bag system that deployed, were compiled into a subset of the NHTSA EDR dataset. A total of 132 cases met these criteria. Using the EDR data, air bag deployment times were used to form a cumulative distribution, as seen in Figure 16. In this sample of GM vehicles, with complete EDR records and equipped with dual air bags, the th percentile deployment time is 2 msec while the 7 th percentile is 3 msec. 29

40 1% 8% 6% 4% 2% % Figure 16. Cumulative Distribution (%) of Driver First-Stage Air Bag Deployment versus Deployment Time (msec) Delayed Deployments The NHTSA EDR dataset contained 132 cases involving deployment of an advanced dual-stage air bag. Twelve vehicles had driver deployment times recorded by the EDR of 72. msec and longer. Four of the vehicles were CAC. Eight cases were pre-cac vehicles with dual-stage air bags. For each of these vehicles, the NASS and EDR data were reviewed to determine common characteristics. The GM vehicle model year, make, and model for these cases as reported by NASS are presented in Table 16. Table 16. Vehicle Model Year, Make and Model (* = CAC Vehicle) NASS Case Model Year Make Model Number * 23 Chev Avalanche * 24 Chev C/K-series pickup Pont Bonneville/Catalina Chev Caprice/Impala Chev Monte Carlo (FWD) Saturn Ion Saturn Ion Chev Caprice/Impala * 24 GMC C,K,R,V-series P/U Chev Equinox Chev Caprice/Impala * 23 GMC C,K,R,V-series P/U 3

41 For each of these cases, the EDR data was reviewed to determine the driver seat belt status, longitudinal delta V of the case vehicle, and the driver s air bag deployment time. This data is shown in Table 17. Table 17. Driver Belt Status, Vehicle Longitudinal Delta V, and Driver Air Bag Deployment Times (* = CAC Vehicle) NASS Case Number Driver Belt Status Delta Driver Air Bag Deployment Time (msec) * Buckled * Unbuckled Buckled Buckled Buckled Unbuckled Unbuckled Buckled * Unbuckled Buckled Buckled * Unbuckled NASS Case Discussion The following presents a short description of the crash, vehicle speed, and longitudinal delta V as reported by the EDR, multi-event as reported by the NASS investigator or the EDR, and some potential reasons for the long reported driver s air bag deployment times. In the discussions that follow, PDOF refers to the principal direction of force, expressed in degrees, where is direct frontal. GAD refers to the general area of damage. GAD = F indicates frontal damage Impact description: Minor vehicle impact, followed by curb hit (EDR N/D event) and then subsequent vehicle impact (EDR D event). Vehicle speed: -1 sec = 33 mph D event delta V = 6 mph GAD/PDOF: Frontal/3deg Multi-event: yes Potential reasons for late-reported deployment time: Low-delta-V event Closely spaced D and N/D events Impact Description: Multi-event crash sideswiped small post, offset impact on utility pole (D event) followed by curb hit. 31

42 Vehicle speed: -1 sec = 1 mph D event delta V = 19 mph GAD/PDOF: F/deg Multi-event: yes Potential reasons for late-reported deployment time: Extreme low overlap with pole (soft) May miss satellite sensor on lower radiator support Abnormal delta V increases at 1 msec Impact Description: Vehicle front contacted a mailbox and a utility pole and came to rest against the pole. Vehicle speed: -1 sec = 42 mph D event delta V = 29 mph GAD/PDOF: F/deg Multi-event: yes Potential reasons for late-reported deployment time: Narrow impact (soft) Impact Description: Vehicle struck a street sign and a large diameter tree. Vehicle speed: -1 sec = 76 mph D event delta V = 8 mph GAD/PDOF: F/deg Multi-event: yes Potential reasons for late-reported deployment time: Narrow impact (soft) Delayed start of delta V data. No vehicle acceleration from AE to ~4 msec Impact Description: Vehicle departed the left side of the road, hit curb, and contacted a concrete utility pole on the median with its front. Vehicle speed: -1 sec = 48 mph D event delta V = 26 mph GAD/PDOF: F/3deg Multi-event: yes Potential reasons for late-reported deployment time: Narrow impact pole (soft), with broad damage Delta V recording shows no vehicle acceleration from AE to ~3 msec Impact Description: Vehicle struck a wooden utility pole with its front, shearing the pole, which resulted in the vehicle rolling 6 quarter turns. Vehicle speed: -1 sec = 61 mph D event delta V = 11 mph 32

43 GAD/PDOF: F/2deg Multi-event: no (yes subsequent to D event) Potential reasons for late-reported deployment time: Low-delta-V event Narrow offset impact (soft) Impact Description: Vehicle rear-ended stopped vehicle in roadway at stop sign. Vehicle speed: -1 sec = 49 mph D event delta V = 3 mph GAD/PDOF: F/deg Multi-event: no Potential reasons for late-reported deployment time: Broad damage Delta V recording shows no vehicle acceleration from AE to ~2 msec Impact Description: Vehicle struck another vehicle on roadway (sideswipe), struck a fire hydrant with its front plane (D event), and then struck a steel sign pole Vehicle speed: -1 sec = 17 mph D event delta V = 8 mph GAD/PDOF: F/1deg Multi-event: yes Potential reasons for late-reported deployment time: Low-delta-V event Narrow impact Impact Description: Other vehicle swerved to miss debris on roadway and impacted subject vehicle head on with small overlap Vehicle speed: -1 sec = 29 mph D event delta V = 18 mph GAD/PDOF: F/34deg Multi-event: yes Potential reasons for late-reported deployment time: Narrow offset impact May miss satellite sensor on lower radiator support Abnormal delta V increases at 3 msec Delayed start of delta V data. No vehicle acceleration from AE to ~ msec Impact Description: Vehicle contacted a signpost, 2 wooden boxes, another post, and a third wooden box. Vehicle speed: -1 sec = 4 mph 33

44 D event delta V = 8 mph GAD/PDOF: F/deg Multi-event: yes Potential reasons for late-reported deployment time: Low Delta V event Narrow offset impact (soft pliable planter box) Delayed start of delta V data. No vehicle acceleration from AE to ~ 2 msec Impact Description: Other vehicle crossed center and hit subject vehicle with extreme offset engagement. Vehicle speed: -1 sec = 43 mph D event delta V = 8 mph GAD/PDOF: F/deg Multi-event: no Potential reasons for late-reported deployment time: Low-delta-V event Offset to left side Narrow impact May miss satellite sensor near hood latch Velocity change trace starts at 8 mph at 1 msec Impact Description: The right front fender was struck by another vehicle at an intersection followed by the subject vehicle hitting a signal pole Vehicle speed: -1 sec = 19 mph D event delta V = 2 mph GAD/PDOF: F/deg Multi-event: yes Potential reasons for late-reported deployment time: Pole impact (soft) Misses frame rails Offset impact (away from satellite sensor, if equipped) Discussion Abnormal delta V traces: On at least 2 of the 12 cases investigated, the EDR recorded the vehicle s speed increasing during the impact. In case , this was observed. Figure 17 shows this data. 34

45 Figure 17. Case EDR Delta versus Time (msec) Figure 18. Case Differentiated EDR Delta versus Time (msec) A closer examination can be made by differentiating these data to obtain a rather crude representation of the vehicle deceleration. This is shown in Figure 18. From this data there is clear vehicle acceleration at 11 msec. While it is not unusual to see positive acceleration in the high-frequency acceleration data, it is unusual to see it in low-frequency data. Since this data represent very lowfrequency data, an occurrence of this type should be considered abnormal. A review and validation of this process is found in the Appendix. Delayed start of delta V data: In several cases the data captured and recorded is part of the EDR record related to the deployment file shows rather long delays between AE and significant changes in vehicle delta V. An example of this is found in case , where the delay was about milliseconds. Figure 19 shows the first major separation from mph to be at 6 msec. 3

46 Findings Figure 19. Case EDR Delta versus Time (msec) The following is a discussion of these 12 cases. Because this is a very small sample and because case counts are used, rather than weighted data, generally only qualitative statements are made. A review of the model years for these 12 case vehicles shows fairly even distribution, given the small sample and the fact that newer vehicles were not available for selection in the earlier case years. This data is shown below. Table 18. Distribution of Model Years in Delayed Deployment Cases Vehicle MY Number of Cases A review of the vehicle type also shows no trends. Both trucks and passenger vehicles had long recorded driver s air bag deployment times. Also, several GM brands were found in the list, as were various sizes of passenger vehicles. Furthermore, driver seat belt status varied between the cases as did crash severity, ranging from 6 mph to nearly 6 mph. Several common characteristics were found among these 12 cases. Narrow/Offset: In many of the cases, the vehicle hit something narrow, such as a pole. Others had significant offset impacts, typically engaging a small portion of the vehicle. Narrow impacts tend to be softer because they may not involve the frame rails. Figure 2 and Figure 21 present examples of these impacts. 36

47 Figure 2. Case Impact With Small Sign and Pole Figure 21. Case Showing Fire Hydrant Damage on Vehicle s Right 37

48 Low Delta V: Several cases had low-delta-v crashes. These crashes are in the zone where the air bag may or may not deploy. For some of these crashes, more time may be needed for the air bag controller to predict the need for air bags deployment, hence the longer deployment times. Abnormal data: As mentioned in the case description section above, some cases had what might be construed as abnormal or unexpected data. There were at least three categories of abnormal data. Two of these were discussed earlier in the Discussion section. Delayed onset of significant changes in velocity after time zero, also referred to as AE. Reversal in the delta V characteristic High starting point for the delta V trace, as reported at the 1 ms data point. Multi-Impact: Many of these 12 cases have earlier non-deployment impacts, as reported by both NASS and the EDR. Table 19 summarizes these characteristics by NASS case number. Table 19. Summary of Delayed Deployments (* = CAC Vehicle) NASS Case Number Narrow/ Offset Low DV Abnormal data Multi-Impact * * * * Advanced Air Bag Suppression Performance The driver and front-passenger restraints can operate independently in an advanced air bag system. Deployment of the driver air bag does not always imply that the passenger air bag will also be deployed. Deployment of the right-front passenger air bag can be suppressed under certain conditions. A manufacturer may choose, for example, to not deploy the passenger air bag if there is no occupant seated in the right-front passenger location. More importantly, the air bag may be suppressed if a child is detected. Table 2 shows the frequency of non-deployments for right-front passenger air bags in crashes sufficiently severe to deploy the driver frontal air bag. All cases 38

49 in this table involve CAC vehicles. In three of the cases, occupant descriptions were not available as the vehicles were not inspected by NASS investigators. Right-front passengers were present in 14 of the 44 remaining cases. Table 2. Frequency of Right-Front Passenger Air Bag Non-Deployments in Crashes in Which the Driver Air Bag Deployed in CAC Vehicles Right-Front RF Air Bag RF Air Bag Total Passenger Deployed Non-Deployment Adult Child None Total When the passenger seat was vacant, the passenger air bag did not deploy in the majority of the cases (27 of 3). This indicates the presence of sophisticated occupant sensors that are characteristic of advanced air bag systems. This behavior, however, can be dependent on the air bag control module as automakers have the flexibility to implement or not implement this non-safetyrelated feature. Only the SDMFG22 module suppressed the air bag if the passenger seat was vacant (27 of 27). The SDMDW23 module on the other hand deployed the right-front air bag despite the fact that no occupant was seated at that location (3 of 3). We believe that this is the result of how the air bag control module was programmed, rather than an error by the air bag control module. Air Bag Non-Deployment in the Presence of a Right Front Passenger Deployment of the driver air bag does not always imply that the passenger air bag will also be deployed. Table 2 shows two particular cases of interest in which the passenger air bag did not deploy despite the presence of a right-front passenger. In both cases, the driver bag deployed. In both cases, the passengers were subjected to a longitudinal delta V of over 2 mph. Earlier in this paper, these cases were referred to as split deployments. In the first case (NASS/CDS case ), the right-front passenger was a -year-old male child weighing 2 kg. The child was not seated in a child safety seat. The subject vehicle, shown in Figure 22, was a 24 Chevrolet C/K-series pickup truck that struck a guardrail and then suffered a rollover. The EDR recorded a longitudinal delta V of 2.3 mph in the guardrail impact. NASS investigators estimated a PDOF of 3 degrees. The NASS investigator indicated that the child was restrained by a three-point belt. The EDR however recorded that the right-front passenger belt was not buckled. The air bag on/off switch was in the auto position. However, when a child is detected, CAC vehicles are designed to either suppress the air bag or deploy the air bag in a low-risk manner. In this case, the system appears to have detected the child and correctly suppressed the passenger air bag. 39

50 In the second case (NASS/CDS case ), the right-front passenger was a 29-year-old male restrained by a three-point belt. The subject vehicle was a 23 GMC C/K-series pickup truck that was subjected to a frontal crash with a longitudinal delta V of 22 mph at a PDOF of 1 degrees. As with the previous case, three reasons were investigated for air bag nondeployment: air bag on/off switch, failure of weight sensor, and a forward-located seat. NASS investigators noted that the air bag on/off switch was in the auto position. Vehicle interior photos also showed the switch clearly in the auto position. The passenger had a weight of 79 kg and height of 17 cm. There is little chance that a properly functioning weight sensor would not have detected this occupant. The EDR recorded that the passenger seat position was in the rearward position making this also an unlikely reason for air bag suppression. One other possible scenario is that the auto/off switch status was tampered with post-crash. Unfortunately, the EDR data as downloaded with the Vetronix reader only indicates that the right-front passenger air bag was suppressed. The EDR does not indicate whether the nondeployment was due to the auto/off switch being set in the off position or whether the nature of this particular crash did not meet the air bag deployment criteria. (a) Frontal and Rollover Damage to 24 Chevrolet Silverado Subject Vehicle (b) Passenger Air Bag On/Off Switch in Auto Position Figure 22. Frontal Crash Followed by a Rollover in Which Driver Air Bag Deployed, But Passenger Air Bag Did Not Deploy for a Child in the Right-Front Seat (NASS ) 4

51 (a) Frontal Collision Damage of a 23 GMC Sierra Pickup (b) Passenger Air Bag On/Off Switch in Auto Position Figure 23. Frontal Crash in Which Driver Air Bag Deployed, But Passenger Air Bag Did Not in the Presence of an Adult Right-Front Passenger (NASS ) Limitations This study has several limitations: The study was based on a limited dataset of vehicles having advanced air bags. Because of the small sample currently available, the conclusions of this analysis should be regarded only as an initial indication of the more conclusive findings that can be expected from follow-on studies with a larger EDR sample. All vehicles were manufactured by General Motors. The results may not apply to other automakers. The frequency distributions presented in this paper apply only to the study dataset. Because the study has not used NASS/CDS case weights, the results should not be interpreted as necessarily representative of the U.S. national crash environment. 41